So here is a number that stopped me mid sentence during this interview. Less than five percent. That is how often a customer talking to Lenovo's AI native customer support agent gets bumped to a human being. Not five percent of easy questions. Five percent of everything, software problems, hardware problems, policy questions, status checks, even upsells, all lumped together. Jack Zhang, who leads AI transformation and worldwide eSupport strategy for Lenovo, told me on this episode that he considers it something close to an industry record for technical support. Honestly, after looking at where the rest of the industry sits, I believe him.
Jack just took home Executive of the Year honors at the 2025 BIG Awards for Business, and once you hear the full story, the award makes a lot of sense. He has spent 17 years at Lenovo, a company that held the top spot in global PC market share with 17.3 percent year over year growth as of the most recent IDC data, and he is currently in charge of the eSupport platform across the company's ROW region, meaning the support website, social channels, the community forums, the on device app, and the AI chat agents themselves, all more than 180 countries worth of it. That is a lot of surface area for one leader, but it is also exactly why his answers on this episode carry so much weight.
An Eight Year Head Start Most People Don't Know About
Here is the part that surprised me. Lenovo did not wake up in 2023 and decide to bolt some AI onto its support site because everyone else was doing it. Jack traced the company's AI journey back roughly eight years, starting with basic machine learning chatbots, then moving to NLP based bots, then to the generative AI and large language model tools most of us think of as new. What he calls the super agent today is really the fourth or fifth generation of a system Lenovo has been quietly rebuilding for the better part of a decade. Recent industry data backs up why that patience matters. Gartner predicts that 40 percent of enterprise applications will feature task specific AI agents by the end of 2026, up from under five percent in 2025, and Cisco's 2025 survey of nearly 8,000 leaders found that over 56 percent of customer support interactions will use agentic AI by mid 2026. Lenovo was already deep into its second or third rebuild by the time most companies are just starting.
Jack was clear that none of it works without the right data foundation underneath it. He called Lenovo's decades of customer service history and content a treasure, and said the whole project comes down to one blunt question, whether there is enough clean data and content to actually support the AI agent, because garbage in still means garbage out no matter how good the model is. That foundation became a shared knowledge management layer that serves human agents and AI agents at the same time, across every business unit, not a bolt on tool sitting off to the side.
Rebuilding The Process Instead Of Just Copying It Into AI
This is the piece I think most companies get wrong, and Jack said it plainly. You cannot just take your existing human workflow and hand the same steps to an AI agent and call it transformation. He pushed his team to ask a harder question, if Lenovo got to design the support process from scratch today, what would it actually look like. That mindset is really what AI native customer support comes down to. It produced what he calls an AI native organization, one where AI is not managing tickets a human used to manage, it is running an entirely reshaped process alongside a layer of what Jack calls AI teammates, including AI alert agents, AI notification agents, and AI quality assurance agents that simulate and test new versions before anything reaches a real customer.
Jack also made an argument that stuck with me, that the biggest obstacle to this kind of transformation is rarely the technology anymore, it is the organization itself. Lenovo's leadership went further than most companies would, merging the mobile, PC, and server support teams into a single group running on one knowledge platform, rather than leaving three siloed teams to each bolt AI onto their own separate process.
The Metrics That Prove It Is Actually Working
Numbers are where this conversation really earns its Executive of the Year nod. Beyond the sub five percent transfer rate, Jack shared that roughly 60 percent of hardware related interactions are now fully touchless, meaning the AI agent diagnoses the issue, checks spare parts inventory, confirms availability, creates the service order, and dispatches it to a depot center, all without a live agent ever getting involved. Another 20 percent are what he calls partial touchless, where a live agent only needs to click one button to approve a specific step rather than run the whole conversation over again. Combine those two categories and Lenovo is automating close to 80 percent of its support journeys end to end, often in under ten minutes total.
Context matters here. Industry benchmarking on AI support resolution generally puts mature deployments somewhere between 60 and 76 percent resolved, with best in class vendors reaching into the mid 80s on transactional queues, and that is resolution, a looser bar than a full touchless service dispatch. Lenovo's numbers sit at the sharp end of that range while measuring something considerably harder to pull off.
Where This Goes Next
Jack ended our conversation on something that genuinely caught me off guard. His team recently noticed that three to five percent of the survey feedback Lenovo collects is now coming not from a human customer at all, but from that customer's own AI assistant, checking on a support ticket on their behalf. He called it scary in the best possible way, evidence that the interface between company and customer is quietly shifting from a person typing into a chat window to one AI system talking to another. Lenovo is already building toward that world, optimizing content the way companies used to optimize for search engines, except now the audience reading it is another AI model, and building the agent to agent connections that let a customer's assistant work directly with Lenovo's support agent with no human touch at all. It tracks with what customers say they already expect. Zendesk's 2026 CX Trends research found that 74 percent of consumers now expect 24 hour service and 81 percent want an agent to pick up a conversation without making them repeat themselves, which is close to a description of the exact system Jack just described.
If you want the full conversation, including Jack's take on where customer service expectations are headed next and what he wishes more companies understood before they start their own AI rebuild, you can catch the whole episode on The Winners' Circle podcast. Subscribe so you don't miss the next one.